Marijuana's risk to drivers debated

[mike_c]

87. did you read it?

It's somewhat flawed, although that does not necessarily make it wrong-- it just reduces confidence in the results. First and foremost, they used ONLY data from drivers who had crashes resulting in injury or death, and specifically excluded any data from cannabis users who did NOT have accidents. It seems to me that's a test of an altogether different hypothesis than the one the authors published, i.e. people who have accidents are somewhat more likely to have used cannabis than not, but that is certainly NOT the same as establishing a causal relationship between cannabis consumption and impaired operation.

Second, it's a meta-analysis that specifically excluded experimental and simulation data, which tends to provide evidence contrary to the authors' assertions. They excluded a LOT of data, and cherry-picked what they themselves defined as "high quality" data sets based on a priori conditions that they defined. I'm always suspicious about that sort of data selection.

Finally-- and I've read the paper a couple of times, but still find this unclear-- the authors say that they've eliminated bias caused by concurrent alcohol use by only using data about serum or urine THC and cannabidiol concentrations, which are inherently misleading by themselves due to long system retention times without any detectable "impairment," but also because the authors are not clear about how this omission screened out alcohol abuse bias-- it SOUNDS as though they simply ignored any data about concurrent use, rather than using such data to screen out confounding factors. But later they again note that ignoring confounding issues removed them from the analysis, which doesn't make sense, so I hope I'm just misunderstanding that part of their methods.